As the Arctic Warms, AI Predictions Include Moving Sea Ice
For generations, the Arctic residents count on seasonal sea ice, which grows and recedes throughout the year. Polar bears and marine mammals rely on it as a hunting and resting place; Indigenous people fish from ice openings known as polynyas, and use well-known routes through the ice to travel from one place to another. But Arctic air and water have been warming three times faster than the rest of the planet since 1971, according to a May 2021 study. report to the Arctic Council, and this warming causes the ice to expand and shrink in unpredictable ways.
Some scientists and research firms are now deploying artificial intelligence-powered tools to provide more accurate and timely predictions of what parts of the Arctic Ocean will be covered by ice, and when. AI algorithms support existing models that use physics to understand what is happening at the surface of the ocean, a dynamic zone where cold underwater currents meet with strong winds to create floating ice rafts. This information has become more valuable for members of the Arctic tribe, commercial fishermen in places like Alaska, and shipping companies around the world interested in taking shortcuts through open water parts.
Leslie Canavera, CEO of Polarctic, a Lorton, Virginia-based science consulting firm that develops AI-based forecast models, says the uncertain pace of climate change means existing sea ice models have become less accurate. That’s because they are based on environmental processes that move rapidly.
“We don’t have a good understanding of climate change and what’s going on [Arctic] system, ”said Canavera, who belongs to the Yup’ik tribe and grew up in Alaska. “We have statistical modeling, but then you look at a lot of averages. Then you have artificial intelligence, where it can see system trends and learn.
Existing physics -based models capture hundreds of years of scientific records of ice conditions, current meteorological conditions, the speed and location of polar jet streams, the amount of cloud cover, and sea temperature. The models use that data to estimate future ice coverage. But it takes huge amounts of computing power to break down the numbers, and many hours or days to make a forecast using conventional programs.
While AI also requires complex data and a lot of initial computing power, once an algorithm is trained on the right amount and type of data, it can detect patterns in climate conditions faster than in physics -based models, according to Thomas Anderson, a data. British Antarctic Survey scientists have created an AI ice forecast called IceNet. “AI methods can just run thousands of times faster, as we found in our model, IceNet,” Anderson said. “And they also learn automatically. AI couldn’t be smarter. It does not replace physics-based models. I think the future takes advantage of both sources of information. ”
Anderson and his colleagues published their new model of the sea ice forecast in August in the journal Communication in Nature. IceNet uses a form of AI called deep learning (also used to automate credit card fraud detection, operate self-driving cars, and run personal digital assistants) to train oneself to give a six monthly forecast of each 25-kilometer square grid across the region, based on Arctic climate simulations between the years 1850 to 2100 and actual observational data recorded from 1979 to 2011. When the model was trained and given current meteorological and ocean conditions, IceNet has overcome a leading physics-based model to make seasonal predictions about the presence or absence of sea ice in each grid square, especially during summer, when the ice goes through an annual retreat, according to NATURE study.